Tracing Scribal Variants and Textual Transmission: A Paleographic Approach to the Nanatsu-dera Manuscript of the Dafangguang Rulai Xingqi Weimizang Jing

Abstract

This paper examines the Nanatsu-dera manuscript of the Dafangguang Rulai Xingqi Weimizang Jing (RXWJ) through the lens of scribal practices, with a focus on variant characters (yitizi, 異體字) and textual transmission. As a “separately produced scripture” (bie sheng jing, 別生經), the RXWJ was not included in the woodblock-printed editions of the Chinese Buddhist canon, which limited its circulation and made manuscript copies—such as the Nanatsu-dera manuscript—critical for reconstructing its textual evolution, transmission, and scribal modifications. A detailed paleographic investigation reveals scribal variants, orthographic fluidity, and phonetic substitutions, illustrating both intentional adaptations and unintentional errors in textual transmission. Comparative analysis with Dunhuang fragments and the Taishō Canon further contextualizes these variations, shedding light on the interpretive challenges scribes and readers face. The findings suggest that the Nanatsu-dera manuscript underwent three stages of transmission: (1) it originated from the Fifty-Fascicle edition circulating in China, (2) it was used as a base text (diben, 底本) for manuscript copying in Japan, and (3) it was subsequently re-copied and preliminarily collated by Japanese scribes. By tracing scribal variants and textual transmission through a paleographic approach, this research underscores the critical role of manuscript culture in preserving texts outside the canonical tradition, offering new insights into the mechanisms of Buddhist textual transmission and adaptation in medieval East Asia.

1. Introduction

The Dafangguang Rulai Xingqi Weimizang Jing (大方廣如來性起微密藏經, RXWJ, The Vast and Expansive Tathāgata’s Nature Origination and Subtle Treasury Sūtra) has long attracted scholarly attention due to its profound connection with the thirty-second chapter of the Sixty-Fascicle edition of Huayan Jing (華嚴經, Avataṃsaka Sūtra), translated by Buddhabhadra (佛陀跋陀羅, 359–429 CE; T 278), titled Baowang Rulai Xingqi Pin (寶王如來性起品, BRXP, The Nature-Origination of the Jewel King Tathāgata). The RXWJ was first referenced in the Chusanzang Jiji (出三藏記集, A Collection of Records on the Translation of the Tripiṭaka), compiled by Sengyou (僧佑, 445–518 CE) in ca. 515 CE. It falls into the category of “translator unknown” (shiyi, 失譯; T 2145, 21c18). However, the Kaiyuan Shijiao Lu (開元釋教錄, KSL, Catalogue of Śākyamuni’s Teachings Compiled during the Kaiyuan Era), compiled by Zhisheng (智昇) in 730 CE, compared it to the BRXP and determined that their contents were not significantly different. As a result, Zhisheng categorized the RXWJ as a “separately produced scripture” (bie sheng jing, 別生經; T 2154, 652b15), emphasizing its status as an independent adaptation rather than a direct translation.1 This classification was particularly significant as the KSL served as a foundational reference for determining which texts were included in the woodblock-printed editions of the Chinese Buddhist canon. The designation, which excluded the RXWJ from the KSL’s catalog of canonical scriptures (ru zang lu 入藏錄)2, consequently limited its transmission and resulted in its absence from the woodblock-printed editions of the Chinese Buddhist canon.

For much of the 20th century, in the absence of access to the RXWJ itself, scholars were confined to analyzing its title and scattered references. As summarized by Kaginushi, this includes Tokiwa Daijō 常盤大定 (1938), Takamine Ryōshū 高峯了州 (1942), Nishio Kyōyū 西尾京雄 (1953), Kagawa Takao 香川孝雄 (1954), Sakamoto Yukio 坂本幸男 (1955), and Ishii Kyōdō 石井教道 (1967) (Kaginushi 1973, pp. 37–38). Their research primarily examined its potential relationship with the BRXP and another corresponding sūtra, the Rulai Xingxian Jing (如來興顯經, The Manifestation of the Tathāgata; T 291), translated by Zhu Fahu (竺法護, Dharmarakṣa, 239–316 CE). For instance, Kaginushi analyzed the term xingqi (性起, “nature manifestation”), suggesting that the character xing 性 was a later addition by Buddhabhadra or his disciples (Kaginushi 1974, pp. 313–14). However, these studies were constrained by insufficient evidence as the primary text was unavailable.

The discovery of the Nanatsu-dera Manuscripts has significantly reshaped this discourse by providing a nearly complete version of the RXWJ, enabling direct textual analysis. Kimura’s critical edition of RXWJ is included in The Long Hidden Scriptures of Nanatsudera Research Series IV (七寺古逸經典叢書第四卷, Makita and Ochiai 1999), which also features editions of other sutras prepared by various scholars. This edition consists of five sections: (1) Guidelines (Hanrei, 凡例), (2) Photographic reproduction (hereafter referred to as Nanatsu-dera’s RXWJ) with a transcription of the manuscript (Kimura’s Transcription, as shown in Figure 1), (3) Collation (Kōchū, 校注), (4) Commentaries (Chūki, 注記), and (5) Combined Annotation (Kaidai, 解題). In the section on Collation, Kimura presents information derived from the Taishō Shinshū Daizōkyō (大正新脩大正藏: hereafter, Taishō Canon), which collates the Huayan Jing with scriptures in the Shōgozō (聖語藏)3 and other printed editions of the Buddhist Canon.

Despite these contributions, Kimura’s analysis of the variant characters (yitizi, 異體字) is relatively brief, leaving their cultural and textual significance underexplored. For example, Kimura identifies various omissions and discrepancies in both characters and phrasing in Nanatsu-dera’s RXWJ, but he does not explore the underlying reasons for these textual phenomena. These limitations hinder a deeper understanding of the historical context and the dating of the original Chinese manuscript from which the Old Buddhist Manuscripts in Japan (日本古寫經) were transcribed. Scholars have noted a close relationship between the Nanatsu-dera manuscripts and early Chinese Buddhist texts. For example, Liang points out that some scriptures in the collection preserve content lost in China between the third and fifth centuries. She suggests that the Nanatsu-dera versions likely entered Japan early and are closely related to the Nara-period manuscript lineage (Liang 2003, p. 29). As a result, the Nanatsu-dera manuscript’s potential as a preserved writing feature of Buddhist texts from the Sui-Tang period (隋唐時期, 581–907 CE) and its similarities to the Dunhuang manuscripts remain insufficiently evaluated.

This paper introduces the Nanatsu-dera manuscript and Kimura’s edition, laying the groundwork for analyzing its textual features. It then investigates these features using codicological and paleographic methods, informed by recent research on Dunhuang manuscripts. Codicologically, the paper follows Kimura’s observations and considers the manuscript’s characters-per-line format to reflect its physical layout and transmission context. Paleographically, it examines handwriting differences, character forms, and possible causes of scribal errors; text-critically, the manuscript’s readings are analyzed in comparison with two Dunhuang fragments of the Huayan Jing (BRXP) to identify textual variants. Moving beyond a binary of “correct” and “incorrect”, this paper focuses on writing features and their historical implications. By tracing scribal variants, it argues that Nanatsu-dera’s RXWJ reflects a three-stage transmission process and underscores the manuscript’s analytical value—particularly its contribution to understanding the textual transmission of the Huayan Jing.

2. Codicological Features of Nanatsu-dera’s RXWJ

Nanatsu-dera’s RXWJ consists of two fascicles, both in accordion-fold format. The database Old Buddhist Manuscripts in Japanese Collections (日本古写経データベース) provides online access to the initial part (one page) of the manuscripts for viewing. As shown in Figure 2, the outer title, inner title, and tail title have been all well preserved. However, the beginning section of the second fascicle shows some damage, with approximately eight lines of text either completely or partially illegible.

Kimura observed that the two fascicles of Nanatsu-dera’s RXWJ share consistent physical characteristics. The first fascicle comprises 32 pages, each measuring 25.9 cm in height and 53.2 cm in width, with 27 lines per page (based on the second page as a reference). The top and bottom margins are marked in vermilion, while the vertical dividing lines are drawn in ink. The top margin measures 2.0 cm, and the bottom margin is 3.4 cm. Each line is spaced 20.4 cm apart, with a column width of 1.9 cm. Most lines contain 17 characters, with slight variations. The second fascicle exhibits nearly identical dimensions but consists of 25 pages instead of 32 (Kimura 1999, p. 675).

As Kimura points out, the lines in Nanatsu-dera’s RXWJ generally consist of 17 characters, though variations exist. The number of characters per line generally ranges from 15 to 20, with 16 to 18 being the most common. Notably, the font size at the beginning of each line is often larger, whereas the last 3–5 characters are written smaller, with the spacing between them gradually tightening toward the bottom. As illustrated in Figure 3, the top six characters, marked with red squares, highlight the larger font size and wider spacing. In contrast, the bottom six characters show a gradual reduction in font size and tightening of spacing as the line progresses. These adjustments suggest that while there was no rigid standard for the number of characters per line, the base text (diben, 底本) likely aimed for approximately 17 characters.

Furthermore, as shown in

Table 1. Five instances in Nanatsu-dera’s RXWJ of omitted strings of words.

Kimura’s Transcription	BRXP in the Taishō Canon	Omitted Strings of Words
生育眾生未曾覩見。何以故。無肉之所饒益 (Kimura 1999, p. 585)	生盲眾生未曾覩見。何以故。無肉眼故。佛子。此生盲眾生雖不見日。亦為日光之所饒益。 (T 278, 616c1-3)	眼故佛子此生盲眾生雖不見日亦為日光 (17 characters)
妙香色雲出種種衣雲、出種種雜色雲4光 (Kimura 1999, p. 606)	妙香色雲。出種種衣色電光。種種衣色雲。出妙香色電光。淨水色雲。出種種雜色電光。 (T 278, 620b09-11)	色電光種種衣色雲出妙香色電光淨水色 (17 characters)
佛子、於意云何、彼大海水、為如是海水、深廣無量 (Kimura 1999, p. 635)	佛子。於意云何。彼大海水。爲無量不。答言實爾。其水深廣不可爲諭。佛子。如是海水深廣無量。 (T 278, 625c10-13)	無量不答言實爾其水深廣不可為諭佛子 (17 characters)
法界無身故。如來。佛子、譬如鳥飛虚空 (Kimura 1999, p. 637)	法界無身故。如來行亦如是。行亦如是。無量無縛。何以故。如來行無身故。佛子。譬如鳥飛虚空。 (T 278, 626 a 19-21)	行亦如是無量無縛何以故如來行無身故 (17 characters)
最故離放逸 一心常奉持 (Kimura 1999, p. 665)	最勝歡喜衆  此經爲内藏 能出生無量  一切白淨道 是故離放逸  一心常奉持 (T 278, 634b3-5)	勝歡喜衆 此經爲内藏能出生無量 一切白淨道是 (20 characters)

, there are four instances in Nanatsu-dera’s RXWJ where strings of exactly 17 characters have been omitted. These omissions provide strong evidence that the base text may have been organized with 17 characters per line. Such errors likely occurred when scribes inadvertently skipped a line during the manual copying process, a mistake more likely to happen when adjacent lines have similar character counts. Additionally, unlike the prose formatting of these examples, there is a notable instance of scribal misalignment in a verse where each line contains 10 characters. In this case, 20 characters were omitted due to a line-skipping error, further demonstrating how the uniformity of character counts could result in transcription mistakes, even in poetic or structured formats.

The colophon at the end of two fascicles credits Renjōbō (蓮定房, lit. “Lotus-Determination Abode,”), whose dates of birth and death remain unknown, as both the “Scribe” (Hisshi, 筆師) and “Proofreader” (Ikkō-ryō, 一校了, “first proofreading completed”) (as shown in Figure 4), highlighting his dual contributions to the transcription and verification of the manuscript. According to Kimura, Renjōbō was involved in approximately 50 other manuscripts, underscoring his significant role in the production of the Nanatsu-dera Manuscript Collection (Shichitera Issaikyō, 七寺一切經). Based on the script style, calligraphic features, and binding techniques, Kimura dates this manuscript to the late Heian period, specifically the latter half of the 12th century (Kimura 1999, p. 675). This timeline aligns with the era of printed editions of the Chinese Buddhist Canon. However, Kimura does not provide a detailed analysis of how he determined the date of Nanatsu-dera’s RXWJ.

A distinctive codicological feature of Nanatsu-dera’s RXWJ is its use of a “27 lines per page” layout, with most lines containing 17 characters—closely resembling the standardized formatting practices found in Dunhuang manuscripts. Zheng (2021), in his study of the Dunhuang manuscripts of the Huayan Jing, reflects on the standardization practices in early Buddhist scriptural transcription. He explains:

In the early stages of Chinese scriptural copying, while certain general formats were followed, standardized practices had not yet been firmly established. Variations in paper production across different regions led to inconsistencies in paper size, resulting in differences in the number of lines per sheet, which ranged from 22 to 31. Additionally, the number of characters per line was not uniform, varying from 14 to 22 characters. However, after the Sui dynasty, the formatting of manuscripts gradually became more standardized, especially for official manuscripts intended for inclusion in the Buddhist Canon (Tripitaka). These texts adopted a fixed format of 17 characters per line and 26–28 lines per sheet. This standardization not only facilitated proofreading but also made it easier to calculate the total number of characters for transcription.

(Zheng 2021, p. 34); (my translation)

Further evidence of this standardization process is found in several Dunhuang manuscripts. For example, manuscript BD04332, containing fragments of the fortieth fascicle of the Huayan Jing and dedicated by Yang Fa-Zhong (楊法仲), features 25 lines per page with 18–22 characters per line and dates to around the fifth century. Other manuscripts dedicated by Bhikshu Seng Dao-Xiang (僧道祥) display more standardized formats: S.1651, Fascicle 4 (27 lines per page, 17 characters per line); S.1608, Fascicle 33 (30 lines per page, 17 characters per line); and BD14438, Fascicle 36 (26 lines per page, 17 characters per line). Although the precise dates of these manuscripts remain uncertain, based on paper quality, layout, calligraphy, and stylistic features, scholars generally attribute them to the sixth century, covering the Southern and Northern Dynasties through the Sui Dynasty (Zheng 2021, p. 38).

These findings corroborate Zhao (2019, p. 168), who points out that the 17-characters-per-line format became the nationally standardized mode for scriptural transcription during the reign of the Southern and Northern Dynasties and was widely adopted in Dunhuang manuscripts. However, as Wu (2024) notes, achieving strict 17-characters-per-line consistency in actual manuscript copying proved challenging. Scribes had to maintain continuous concentration to ensure character accuracy and layout consistency. Under the high-intensity conditions of manuscript copying, even imperial scribes frequently deviated from the standard, resulting in occasional lines exceeding or falling short of 17 characters (Wu 2024, p. 27).5

As Figure 3 illustrates, font size and spacing variations reflect the flexibility scribes employed to approximate the 17-characters-per-line standard. These subtle modifications in Nanatsu-dera’s RXWJ exemplify the inheritance of transcription practices from China to Japan, demonstrating their continuity and adaptation in the historical development of Buddhist manuscript traditions.

The transcription of Buddhist scriptures in Japan can be traced back to 673 CE. However, formal and organized efforts in sutra copying and literary activities only matured during the Nara period (710–794 CE). During this time, government-established transcription centers and temple-based facilities systematically produced Buddhist manuscripts heavily influenced by Chinese traditions. Subsequently, this sutra copying tradition reached its zenith during the Heian period (794–1185 CE), with most transcriptions based on Nara-period manuscripts. Even as late as the Kamakura period (1185–1333 CE), when the Song Dynasty (960–1279 CE)-printed editions of the Tripitaka were gradually introduced to Japan, manuscript versions of the Tripitaka were still primarily based on Nara-period manuscripts (Kajiura 2010, pp. 435–37).6

In this context, scholars such as Fang (2006), Kajiura (2010) and Ochiai (2010) highlight the pivotal role of the Nana-period manuscripts in the history of Japanese manuscript transcription. While it is difficult to confirm that all Nanatsu-dera manuscripts were direct copies of Nara manuscripts, they are widely acknowledged as belonging to the Nara manuscript tradition. These manuscripts represent vital connections in the development of Japanese Buddhist textual culture, reflecting close ties to the original forms of Sui-Tang (581–907 CE) Buddhist texts and notable parallels with the Dunhuang Buddhist manuscripts (Ochiai 2010, pp. 113, 118, 120). Although Kimura dates the manuscript to the late Heian period (Kimura 1999, p. 675), a closer analysis of its variant and erroneous character forms reveals strong similarities with Dunhuang manuscripts, particularly the BRXP fragments. In some cases, even apparent copying mistakes may reflect graphic conventions associated with earlier Chinese scribal practices.

3. Paleographical Characteristics of Nanatsu-dera’s RXWJ

This section examines the paleographical features of Nanatsu-dera’s RXWJ, focusing on two key aspects that shed light on its scribal practices, historical character variants, and connections to Dunhuang manuscripts. The first part reviews Kimura’s (1999) analysis of variant characters (yitizi, 異體字) in his critical edition; the second part incorporates the classification of nonstandard characters (suzi, 俗字) from Dunhuang studies to analyze graphic variations within Nanatsu-dera’s RXWJ.

3.1. Variant Characters in Nanatsu-dera’s RXWJ

Within the “Guidelines” section of Kimura’s critical edition, he provides introductory notes regarding his transcription of Nanatsu-dera’s RXWJ. In item 5, he states, “Variant characters (itaiji, 異體字) and simplified forms (ryakujji, 略字) have been replaced with standard characters (seiji, 正字)”. Although he does not explicitly distinguish between variant characters and simplified forms, he includes a brief list of such characters (Kimura 1999, p. 556). See

Table 2. This table presents the characters Kimura (1999) revised in his transcription of Nanatsu-dera’s RXWJ. It includes examples of variant characters and simplified forms (which I have labeled as “Original Form”) that were replaced with standard characters (labeled as “Correction”), as outlined in the “Guidelines” section.

Original Form	Correction	Original Form	Correction	Original Form	Correction
无	無	尒	爾	囙	因
猒	厭		珍		譬
𠛴	剎	乗	乘	弥	彌
鄣	障		槃		雖
	醫	扵	於		涼
	解	刧	劫	㓛	功
	寂		冥	呪	咒
𢙢	恐	𤍽	熱	蘓	蘇
灾	災		齋

.

Beyond the list provided in the Hanrei (凡例), Kimura employs two distinct approaches to handling variant characters in Nanatsu-dera’s RXWJ: (1) automatically converting them into standard characters and (2) preserving the original form with notes indicating discrepancies from the Taishō Canon. As illustrated in

Table 3. Examples of Kimura’s approaches to variant characters in Nanatsu-dera’s RXWJ.

1. Automatic Revision			2. Preserved with Annotation
Original Character		Revised Character	Original Character		Noted Divergence from Taishō Canon
§ 1		願	§ 11		映蔽→𦴤蔽
§ 2		隨	§ 12		涅壞→沮壞
§ 3		然	§ 13		憧→幢
§ 4		攝	§ 14		群萠→群萌
§ 5		坐	§ 15		生肓→生盲
§ 6		師	§ 16		療治病→治眾病
§ 7		莊	§ 17		柔濡→柔軟
§ 8		明	§ 18		無㝵→無礙
§ 9		軟	§ 19		癊→痰𤸌
§ 10		博	§ 20		無滿→充滿

, the first section provides examples where characters have been revised into standard forms without annotation (Approach 1); the second section presents cases in which the original form is retained and accompanied by explanatory notes (Approach 2).

Obviously, the automatically revised characters (Approach 1) in Kimura’s edition include both variant and simplified forms, whereas his annotations for preserved character forms (Approach 2) lack detailed examination. These alternative written forms often reflect regional, temporal, and stylistic variations, offering valuable insights into the linguistic and historical context of the manuscript. Building on this foundation, analyzing the variant written forms of Chinese characters in Nanatsu-dera’s RXWJ is essential. These variants typically arise through the addition, omission, substitution, rearrangement, or modification of character components—changes frequently influenced by writing habits and other contextual factors. Moreover, the findings of Dunhuang’s studies on the phenomenon of variant character usage provide a useful comparative framework for deeper textual investigation.

3.2. Tracing Allographic Parallels in Nanatsu-dera’s RXWJ and Dunhuang Texts

Various terms and classifications exist for the different written forms of Chinese characters. For example, Galambos notes that the terminology for variant characters in Chinese tradition varies across different fields of study. In epigraphy, they are called biezi 別字 (“different characters”); in Dunhuang studies, suzi 俗字 (“popular forms”); and in printed culture, terms such as ezi 訛字, wuzi 誤字, and cuozi 錯字, which are used to describe “erroneous characters”. Modern usage often adopts the broader term yitizi 異體字 (“characters with different forms”) or, less commonly, yigouzi 異構字 (“characters with different structures”). All these terms are defined in contrast to a standard or correct form (zhengzi, 正字), emphasizing their status as deviations from the norm (Galambos 2011, p. 399; 2020, p. VIII).

In this discussion, I also follow the terminology from the English translation of Qiu ([1988] 2000) Wenzixue Gaiyao 文字學概要, referring to yitizi as “allographs”. Qiu defines allographs as characters with the same pronunciation and meaning but different forms and categorizes them into eight types, including variants with added or omitted components (e.g., qie 匧/篋, “a kind of box”), semantographs vs. phonograms (e.g., lei 淚/泪, “tear”), phonetic or positional differences (e.g., xiang 響/响, “make a sound”; qi 棊/棋, “chess”), and corruption or simplification (e.g., zhou 呪/咒, “incantation”) (Qiu [1988] 2000, pp. 299–301). Although Qiu acknowledges classification ambiguities, his framework helps analyze structural and semantic relationships in the Chinese script. For example, the case of “呪” (zhou) and “咒” (also found in Table 2) may appear to show a simple positional shift of components; however, their actual development follows a progression: “祝” (zhu, “to bless”) transformed into “呪”, and 呪 further underwent orthographic corruption, ultimately becoming “咒”. In ancient texts, “呪” was more commonly used, whereas today, “咒” is the predominant form.7 This character is still preserved in the Taishō Canon in the form of “呪” while Kimura (1999) consistently transcribes it as “咒”.8

With the development of research on the classification systems for allographs, a distinction has emerged between structural variants (yigou, 異構) and graphical variants (yixie, 異寫). Structural variants involve changes to the overall structure of a character, such as “淚” (lei, “tears”) becoming “泪” or “災” (zai, “disaster”) becoming “灾” (as shown in Table 2). These changes in radicals or components, often based on the character’s pronunciation and meaning, are relatively limited and easy to identify and analyze. In contrast, graphical variants arise from changes in the strokes of radicals or components, typically resulting from handwriting differences or stylistic variations. For instance, “亞” (ya, “Asia”) may be written as “亜”, “楷” (kai, “standard script”) as “揩”, or “隨” (sui, “follow”) as “随”. These examples illustrate how graphical variants can differ; a single character can generate many such variants (Cai 2011, p. 7).

Dunhuang studies provide a broad foundation for manuscript research, offering valuable insights into textual variations. Research on nonstandard characters from Dunhuang manuscripts has been instrumental in studying allographs in the Nanatsu-dera manuscripts. Wu’s work, Study of the Characters in the Seven-Temple Edition of Xuanying’s Yinyi (Qisi ben Xuanying Yinyi wenzi yanjiu, 七寺本《玄應音義》文字研究; Wu’s original English title), reviews prior research on the classification of variant characters and introduces his own classification of nine types (Wu 2021, pp. 51–107). Wu treated the Dunhuang manuscripts as synchronic textual material, using character components as units of analysis to infer and compare the similarities and differences between the allographs in Nanatsu-dera’s Xuanying Yinyi 玄應音義 and the Dunhuang variants. He concluded that the structural features of characters in the Dunhuang manuscripts and the Nanatsu-dera are highly similar (Wu 2021, p. 135).

For example, as shown in Table 2, the character “” (zhen), originally composed of “玉” (a semantograph) and “㐱” (a phonogram), underwent a transformation in which its phonetic component was replaced by a semantic component, a process classified as substitution (gaihuan, 改換); the standard character “槃” (pan), originally composed of 木 (semantic component) and “般”(phonetic component), is frequently rearranged in Dunhuang manuscripts, with the “舟” component shifted to the left, resulting in the modified form “” (Cai 2002, pp. 167, 170). Additionally, characters like “願” (yuan), originally composed of “頁” (a semantograph) and “原” (a phonogram), underwent intra-character assimilation (zi nei leihua, 字內類化), written as “” (Table 3, § 1) due to the influence of its right component. Similarly, “映” (yin), originally composed of “日” (a semantograph) and “央” (a phonogram), was modified by the adjacent character “蔽” (bi), leading to the addition of “艹” (see Table 3, § 11), an example of top–bottom character assimilation (shang xia zi leihua, 上下字類化).

In harmony with Wu’s observation, the most common types of variant characters found in these manuscripts are simplification (jiansheng, 簡省), symbolization (jihuahua, 記號化), complexification (zengfan, 增繁), dissimilation (yihua, 異化), and orthographic confusion (e hun, 訛混; literally, “mistakenly mixed”) (Wu 2021, p. 162). These features are rooted in the shared historical context of character development and provide a strong foundation for further analysis. I have chosen to focus on three types of phenomena within Nanatsu-dera’s RXWJ that can be effectively explained through the framework of Dunhuang studies and are directly related to the meaning of the characters. Given this focus, the Chinese characters listed in Table 2, which Kimura classified as “variant characters (itaiji, 異體字)” and “simplified forms (ryakujji, 略字),” can be re-categorized within the context of Dunhuang studies. Below, specific components from Table 2 and Table 3 are re-examined, utilizing the classifications of “Simplification or Ancient Characters” and “Orthographic Confusion” to highlight and explain the allographs within Nanatsu-dera’s RXWJ and to infer their potential links to Dunhuang manuscripts.

3.2.1. Case Study: Simplification or Ancient Forms in Nanatsu-dera’s RXWJ

A simplified character form (shengxing 省形; jiansheng 簡省) refers to a reduction in or simplification of radicals or components (Cai 2002, p. 165; Wu 2021, pp. 58–59). This concept corresponds to Kimura’s term “ryakujji 略字”. The variants of “隨” (sui), “然“ (ran), and “攝” (she) presented in Table 3 (§ 2–4) serve as typical examples of simplified forms. There are also numerous examples where “口” (kou, “mouth”) replaces “人” (ren, “person”), such as with “坐” (zuo; § 5) being written as “”.

However, characters such as “无” (wu, 無), “猒” (yan, 厭), “蘓” (su, 蘇), and “囙” (yin, 因) in Table 2, and “” (shi, 師) in Table 3 (§ 6), which appear as simplified variants, can be traced back to ancient scripts, particularly the clerical script (lishu, 隸書) (Wu 2021, pp. 53, 57). This calligraphic characteristic is significant for identifying manuscripts written in the period from the fifth to the sixth centuries.9

3.2.2. Case Study: Orthographic Confusion in Nanatsu-dera’s RXWJ

In Table 2, “刧” (jie, 劫) and “㓛” (gong, 功) also may appear to be cases of stroke reduction or the simplification of standard characters but are actually the results of writing errors caused by the similarity of their components “力” (li) and “刀” (dao). Similarly, “於” (yu) is often mistakenly written as “扵” (wu/yu) because, during the quick writing of “方” (fang), its folds are frequently straightened, resembling “扌” (shou). Alternatively, the character “涅” (nie) also appears in a variant form as “”, which closely resembles the writing of “沮” (ju). As a result, the two characters are confused (as shown in Table 3, § 12).

This phenomenon is very common in Dunhuang manuscripts, where it has been classified as a type of “mixed usage” (hunyong 混用) or “orthographic confusion” (e hun訛混) (Wu 2021, p. 80). The similarity of components or writing habits causes many types of orthographic confusion. Some of these are easy to detect while others closely resemble another standard character. Consider the character “莊” (zhuang) as an example (§ 7). Originally composed of “艹“ (cao), “爿” (pan), and “士” (shi), “莊” often underwent changes during rapid writing. In this process, “爿” was frequently written as “丬” (qiang), which then connected with “艹”, resulting in “疒” (ne) and leading to the formation of “” as a variant (Cai 2002, p. 162). This variant differs significantly from the standard form, but a simplified version, “庄”, has become the modern usage. Kimura transcribed “” as “莊” in his work without providing any annotation.

However, in certain instances, minor alterations to components cause the character to be identified as a completely different word. For example, the character “幢” (zhuang, “banner”; Table 3, § 13) is frequently miswritten as “憧” (chong, “longing” or “yearning”) in Dunhuang manuscripts as a result of orthographic confusion. This variation arises from the component “巾” (jin, “cloth” or “towel”) being commonly miswritten as “忄” (xin, “heart” or the radical for “emotion”). The components “日” (ri, “sun”), “月” (yue, “moon”), and “目” (mu, “eye”) are frequently interchanged (Wu 2021, p. 82), leading to cases where 萌 (meng) appears as “萠” (peng), “盲” (mang) resembles “肓” (huang), and “明” (ming) is found as “眀” (ming) (as shown in Table 3, § 8, pp. 14–15).

Wu states that this mixed type of variation does not hinder reading comprehension within the given context (Wu 2021, p. 80). This is partially accurate. For example, the character “療” (liao, “to treat” or “to heal”) in the phrase “療治眾病” (liao zhi zhong bing, “to treat and cure various illnesses”) appears as ““ (T 278, 617b12; see Table 3, § 16). Although the exact shape may be unfamiliar, the meaning remains clear from context, allowing us to infer it as a variant of “療”. However, subtle interpretive nuances may be overlooked when standardized printed editions are used without attention to variant forms. The broader issue is that graphical variants, though not necessarily impeding comprehension, can meaningfully influence how meaning is distributed and perceived within the text.

3.3. Graphical Variation in Nanatsu-dera’s RXWJ and Its Interpretive Implications

The contents of the BRXP in the Taishō Canon and Nanatsu-dera’s RXWJ are largely consistent; however, the variant character forms preserved in Nanatsu-dera’s RXWJ offer glimpses into how scribal variation might influence textual interpretation. These instances suggest that even minor graphical differences may subtly shift emphasis—especially in descriptions of physical action or force. This interpretive possibility can be explored through the lens of semantic profiling, a concept from cognitive linguistics that explains how varying character forms may foreground distinct facets of a word’s conceptual structure.10 A particularly illustrative example is the variation between “博” (bo) and “搏” (bo) (Table 3, §10) in the BRXP of the Taishō Canon.

In Wu’s study on orthographic confusion, he identifies 18 cases of mutual interchange and 10 of unidirectional substitution (Wu 2021, pp. 80–84). However, his list does not include the interchange between the radicals “十” (shi) and “扌” (shou). Such cases do appear in the BRXP. A notable instance involves a metaphor describing the Buddha as “the king of the Garuḍa” (jin chi niao wang 金翅鳥王), who guides beings toward liberation:

佛子！譬如金翅鳥王，飛行虛空，安住虛空，以清淨眼觀察大海龍王宮殿，奮勇猛力，以左右翅博開海水，悉令兩闢，知龍男女有命盡者，而撮取之。

(T 278, 626b1-4; punctuation added by the author.)

O noble son! It is like the golden-winged king of birds (King of Garuḍa), soaring through the sky and abiding in the vast space. With its pure eyes, it observes the palaces of the dragon kings in the great ocean. Then, with great courage and strength, it vigorously spreads its wings, parting the ocean waters in two. Knowing which male and female dragons are nearing the end of their lifespan, it swiftly seizes them.

(my translation)

While “博” typically conveys meanings related to breadth, expansiveness, or acquiring something. However, these connotations do not match the physical action of the golden-winged bird in this passage. Upon closer examination of Nanatsu-dera’s RXWJ, it becomes evident that “博“ in this context is actually an orthographic confusion of “搏” (Table 3, § 10).11 The character “搏” means “to strike”, “to flap”, or “to beat”, thus contributing to a semantic emphasis on exertion and impact, enhancing the physical intensity of the scene.

As a verb compound, “搏開” (bo kai) follows a causative structure in which the action denoted by 搏 (“to strike”) precedes and enables 開 (“to part”). This reflects a temporal and logical sequence: the striking motion must occur first to achieve the effect of parting. Rather than simply “parting” the sea, the Garuḍa strikes the surface—forcefully beating the waters to split them. Thus, graphical variation influences how the action is perceived, shifting the reader’s understanding of movement and intensity. Jizang’s 吉藏 (549–623 CE) Fahua Xuan Lun (法華玄論, Profound Commentary on the Lotus Sutra) reinforces this interpretive sequence. He writes:

如《華嚴》云：「金翅鳥王上昇虛空，以清淨眼觀大海龍應命盡者，即以兩翅搏水，令兩闢，而撮取之。」

(T 1720, 447b6–8; punctuation added by the author.)

As the Huayan Jing states, “The golden-winged king of birds ascends into the sky. With its pure eyes, it observes the dragons in the great ocean whose lifespans are nearing their end. Immediately, it strikes the water with its wings, causing it to part in two, and seizes them”.

(my translation)

Jizang presents the action as a three-part sequence: striking the water (bo shui 搏水), causing it to part (ling liang pi 令兩闢), and seizing the dragons (cuo qu zhi 撮取之). This structure reflects causative and temporal logic, reinforcing the interpretive significance of “搏” in constructing a vivid and dynamic scene. Based on his phrasing, it can be inferred that Jizang read the character as “搏”, which supports the view that this form was earlier or more semantically precise, with “博” possibly emerging later in the transmission process.

Although “博開” may not cause misinterpretation, the absence or substitution of “搏” (bo, “to strike”) tends to shift interpretive focus from the exertive action to its resulting state. For instance, Śikṣānanda’s translation employs the phrase “鼓揚海水” (gu yang hai shui; T 279, 274c2), which emphasizes the outcome—“churning the ocean”—rather than the physical motion itself. This interpretive divergence is mirrored in exegetical traditions: Fazang 法藏 (643–712 CE), commenting on T 278, refers to the metaphor as jin chi bo hai yu (金翅搏海喻, “the metaphor of the Golden-Winged Bird striking the sea”; T 1733, 412b28), highlighting the action. Chengguan 澄觀 (738–839 CE), commenting on T 279, names it jin chi pi hai yu (金翅闢海喻, “the metaphor of the Golden-Winged Bird parting the sea”; T 1736, 150a4), focusing instead on the result. These terminological choices reflect distinct semantic orientations.

In contrast, compound verbs like 搏取 (bo qu), 博取 (bo qu), or 搏撮 (bo cuo) demonstrate synonymous compounds, where both components jointly express the idea of “to seize”. For example, Śikṣānanda’s version uses 搏取 (bo qu; T 279, 274c3) while Buddhabhadra’ translation presents 撮取 (cuo qu; T 278, 626b4). In such cases, the choice of 搏 (bo), 博 (bo), or 撮 (cuo) has minimal interpretive impact as these compounds are already lexically fused.

To put it succinctly, the example discussed here underscores how paleographic variation can influence semantic focus. The limitations of this case are also acknowledged—similar forms appear in other manuscripts, and thus, it may not establish the distinct textual–critical value of the Nanatsu-dera RXWJ. Nevertheless, it highlights the importance of comparative manuscript analysis and demonstrates how a semantic profiling approach can recover interpretive nuances that standardized editions may obscure.

4. Nanatsu-dera’s RXWJ and the Fragments of BRXP in Dunhuang Manuscripts

Since Nanatsu-dera’s RXWJ closely resembles Buddhabhadra’s translation of the Huayan Jing (T278), Kimura (1999, pp. 679–81) argues that it was established in direct association with the Sixty-Fascicle edition of the Huayan Jing. However, given the inconsistent records regarding this sutra across various Buddhist categories, a crucial question remains: was Nanatsu-dera’s RXWJ a faithful reproduction of an earlier textual tradition, or did it undergo substantial revisions over time? Kimura (1999, p. 681) hypothesizes that the RXWJ was initially compiled several decades after Buddhabhadra’s translation in 420 CE through the extraction of the BRXP. 12 This early version, which Sengyou is known to have referenced, lacked the structural elements necessary for an independent scripture. By the mid-seventh century, a revised version was created to further promote the philosophy of xingqi (性起, “nature manifestation”). However, this revised version did not achieve widespread circulation and was overlooked by Zhisheng when compiling the Kaiyuan Shijiao Lu (KSL) in 730 CE. The surviving RXWJ manuscript from Nanatsu-dera is generally considered a transcription derived from this mid-seventh-century revised version. Specifically, Kimura states,

By 420 CE, the translation of the sixty-fascicle Huayan Jing was completed. In the following year, revisions were made before it began circulating. Over time, particularly by the late fifth century, some individuals took a special interest in the Xingqi Pin (性起品) and extracted it as an independent scripture, which they then compiled and proclaimed as the Weimizang Jing (微密藏經).

(Kimura 1999, p. 679); (my translation)

Kimura’s hypothesis provides an essential foundation for understanding the textual evolution of the RXWJ. His approach, however, does not examine the codicological and paleographic features of the manuscript. Recent research (Zhang and Fu 2014; Zheng 2019, 2021; Zhuang 2019) indicates that several Dunhuang fragments of the Huayan Jing correspond to an earlier edition of Buddhabhadra’s translation, which was later recompiled into the more widely recognized Sixty-Fascicle edition.

Within this textual context, the Nanatsu-dera manuscript emerges as a noteworthy witness to the transmission of the RXWJ. Nevertheless, as far as I can determine, existing studies have not directly compared the Dunhuang fragments with Nanatsu-dera’s RXWJ. A comparative analysis of its scribal conventions alongside the pieces of the BRXP reveals that numerous character forms in Nanatsu-dera’s RXWJ resemble those in this earlier edition, suggesting a possible textual lineage between them.

4.1. The Fragments of BRXP in Dunhuang Manuscripts

As Zheng (2019, p. 7) observes, Chu Sanzang Jiji (出三藏記集; T 2145) describes Buddhabhadra’s translation of Huayan Jing as a “fifty-fascicle edition”, with no mention of a Sixty-Fascicle edition or any evidence of textual reorganization at that time. It was not until the Sui dynasty that Buddhist catalogues, such as Zhongjing Mulu (眾經目錄, A Catalogue of All Scriptures, 594 CE; T 2146) and Lidaisanbao Ji (歷代三寶記, A Record of the Three Treasures Throughout the Successive Dynasties, 664 CE; T 2034), began documenting the existence of a Sixty-Fascicle edition. This trend continued into the Tang dynasty, where catalogues such as Datang Neidian Lu (大唐內典錄, Great Tang Inner Canonical Record, 664 CE; T 2149) and KSL (T 2154) recorded the Sixty-Fascicle edition while simultaneously noting the continued presence of a Fifty-Fascicle edition (Zheng 2019, pp. 6-7). Significantly, evidence from the Dunhuang manuscripts suggests that before the sixth century, during the reign of the Northern and Southern dynasties (420–589 CE) and the Sui dynasty (581–618 CE), Buddhabhadra’s translation of Huayan Jing does not appear to have circulated in a Sixty-Fascicle edition.

Hamar (2013, pp. 91–100) previously compared the Dunhuang manuscripts of the BRXP with the Taishō Canon, documenting textual variations among these sources; however, his analysis was limited to only four manuscripts. In contrast, Zhuang (2019) identifies at least 25 fragments within the Dunhuang materials of the Fifty-Fascicle Huayan Jing that can be attributed to the BRXP.13 Among these, one fragment exhibits a writing style that closely resembles the script found in the Nanatsu-dera manuscript—Nakakura 23.

4.1.1. The Fragment Nakakura 23

In Nakamura Fusetsu’s (中村不折) collection, there are two fragments that Nakamura notes as belonging to the Turfan manuscripts (Tulufan xieben, 吐魯番寫本) (Nakamura 2003, pp. 38, 64). These fragments, known as Nakakura 13 and Nakakura 23, include details such as fascicle numbers, dates, locations, and scribes, allowing their transcription periods to be determined with notable precision. Zheng (2021, p. 32) examines the manuscript’s content and identifies both fragments as early scriptural copies of the Fifty-Fascicle Huayan Jing. As shown in Figure 5, the colophon of Nakakura 23 attributes the text to Fascicle 29 of the Huayan Jing, which corresponds to Fascicle 34 in the Sixty-Fascicle edition. The colophon records the following inscription:

梁普通四年 太歲卯四月 正法无盡藏寫

(Liang Putong si nian taisui mao si yue Zhengfa Wujin Zang xie, “In the fourth year of the Putong era of the Liang dynasty [523], the True Dharma’s Endless Treasury scribe copied this text”.)

Both manuscripts provide compelling textual evidence supporting Sengyou’s bibliographic records, confirming that the Fifty-Fascicle format was the primary structure through which the Huayan Jing was transmitted (Zheng 2021, p. 33). Of the two, Nakakura 23 preserves a larger portion of text and displays features that closely resemble those found in Nanatsu-dera’s RXWJ. The following analysis will, therefore, focus on Nakakura 23 for comparison.

4.1.2. The Scroll of BRXP in the National Museum of China

A further manuscript—previously unexamined in BRXP studies—broadens our view of the Fifty-Fascicle Huayan Jing in the sixth century. It is preserved at the National Museum of China (Zhongguo Guojia Bowuguan 中國國家博物館) and now accessible through the Zhonghua Baodian (中華寶典) Series 6. According to the editor’s introduction, the scroll preserved at the National Museum of China (hereafter referred to as the NMC Scroll) originated from the Library Cave (Cangjing Dong 藏經洞) of the Dunhuang Mogao Caves (Dunhuang Mogao Ku 敦煌莫高窟). The calligraphy follows the distinct “scriptural writing style” (xie jing ti 寫經體), which is characteristic of the Northern Wei period (Bei Wei, 北魏, 386–534 CE) and adheres to a consistent and standardized transcription format (Yang 2022, p. 4). Based on its content and the title at the end, which records it as “Mahāvaipulya Buddhāvataṃsaka Sūtra, Fascicle 29” (Dafangguang Fo Huayan Jing Juan di ershijiu 大方廣佛華嚴經卷第廿九), the manuscript can also be confirmed in its affiliation with the Fifty-Fascicle edition.

Further supporting evidence for its dating is the manuscript’s format and calligraphic style, which are identical to those of National Library 002, suggesting that both were likely transcribed by the same scribe (as shown in Figure 6). As Fang (2022, pp. 1–2) described, the National Library 002 is associated with the Fifty-Fascicle edition. Considering its content, paper quality, script, and calligraphic style, it can be identified as a sixth-century manuscript.

The calligraphic style of the NMC Scroll differs significantly from that of Nakakura 23 and Nanatsu-dera’s RXWJ. As the NMC Scroll has not previously been examined in scholarly literature, its inclusion here is intended to provide a distinct comparative example and to offer new evidence for understanding the early textual history of the Huayan tradition. While my primary focus is on the notable similarities between Nakakura 23 and Nanatsu-dera’s RXWJ, the contrasting features of the NMC Scroll help to further highlight the characteristics shared by the other two manuscripts.

4.2. Comparative Analysis of Scribal Practices

Having introduced the relevant manuscripts, the following section offers a comparative analysis to assess their historical and textual relationships. It is divided into two parts: the first examines codicological features, including line layout and a corresponding textual passage; the second focuses on paleographic characteristics, such as shared and divergent variant graphs.

4.2.1. Codicological Analysis

As noted in the previous Section 2, the omission of entire strings in Nanatsu-dera’s RXWJ suggests that its base text may have been organized into lines of 17 characters. Both Nakakura 23 and the NMC Scroll display this 17-characters-per-line structure as a standard codicological feature. Notably, two omission cases in Nanatsu-dera’s RXWJ—listed as the first and second entries in Table 1—can be plausibly explained as scribal errors. These likely resulted from visual similarity between adjacent lines in the exemplar. Figure 7 presents reconstructed line layouts from Nakakura 23 to illustrate how such errors may have occurred. The red dot marks the omission point corresponding to each example from Nanatsu-dera’s RXWJ.

Figure 7a illustrates an omission involving the characters “无” (wu) and “光” (guang). Due to their visual similarity in clerical script, these characters appear in near-identical positions across adjacent lines, likely causing the scribe to skip a line during copying. Figure 7b presents another example: repeated sequences such as “色” (se), “電光” (dian guang), and “衣色” (yi se) occur at similar positions, again increasing the likelihood of line-skipping. These internal errors suggest that the RXWJ was copied from a source formatted with 17 characters per line.

In addition, the text at the end of the Nakakura 23 fragment—”為道場菩薩 雨大甘露法 隨其所應化 如來心平等” (wei daochang pusa yu da ganlu fa sui qi suo ying hua rulai xin pingdeng)—corresponds exactly to the final portion of the first fascicle of Nanatsu-dera’s RXWJ (as shown in Figure 4 and Figure 5). These codicological features further support the manuscript’s connection to a textual lineage consistent with the Fifty-Fascicle edition of the Huayan Jing.

4.2.2. Paleographic Analysis

Beyond structural similarities, a closer examination of variant characters in these manuscripts provides further insight into the scribal conventions reflected in their production. This section focuses on paleographic features, including shared and divergent forms in Nanatsu-dera’s RXWJ, Nakakura 23, and the NMC Scroll.

A comparison of the characters “魔醯” (moxie), “嚮” (xiang), “增” (zeng), and “停” (ting) across the three manuscripts—shown in

Table 4. Graphic forms in three manuscripts: shared variant characters.

	§ 1	§ 2	§ 3	§ 4	§ 5	§ 6	§ 7	§ 8
Taishō edition	魔醯	響	憎	停	導師	智	礙	痰𤸌
Nanatsu-dera’s RXWJ
Nakakura 23
NMC Scroll	Outside extant portion

(§1–4)—reveals highly similar graph forms. These are cases that Kimura, in his collation of Nanatsu-dera’s RXWJ, identifies as different words due to their graphical distinctions (Kimura 1999, pp. 666–68).

In addition to consistent forms, recurring scribal errors in Nanatsu-dera’s RXWJ provide further clues about its textual lineage. Two common examples are the characters “導” (dao, see Table 4, § 5) and “智” (zhi, § 6), which appear in mistaken or altered forms. A close comparison with Nakakura 23 and the NMC Scroll reveals potential causes. In the case of “導”, the component “寸” (cun) is often written in a highly reduced form. It resembles “小” (xiao) both in shape and size, making it easy for a scribe to misread the character “導” as “道” (dao). For “智”, the component “日” (ri) is sometimes placed beneath the right-hand “口” (kou), a structure seen in early manuscripts but not in the standard script. These graphical similarities point to a visual connection between Nanatsu-dera’s RXWJ and earlier exemplars, suggesting that the scribe may have been working from a model rooted in an older tradition.

Beyond the characters discussed above, several examples in Nanatsu-dera’s RXWJ reflect older scribal conventions that are not visible in printed editions. As shown in Table 3, the character 軟 (ruan) in Nanatsu-dera’s RXWJ manuscript presents two variant forms: 渜 (ruan; Table 3, § 9) and 濡 (ru; § 16); this character is written as “濡” in two sixth-century manuscripts. According to Jiyun (集韻) and Zhengyun (正韻), the character “濡” is pronounced ruan (耎) and shares the same meaning as “輭” (ruan) and “軟 “(ruan), all of which signify softness and pliability. “濡” is also interchangeable with “耎”, further emphasizing its phonetic and semantic interchangeability. Thus, “渜”, “濡”, and “軟” are interconnected through both phonetic and semantic relationships, reinforcing their functional interchangeability in manuscript traditions.

Furthermore, in the phrase describing the Buddha’s constant radiance of boundless, unobstructed wisdom light (常放無量無礙智慧光明 chang fang wu liang wu ai zhi hui guang ming, T 278, 616b19–20), the character “礙” (ai) in Nanatsu-dera’s RXWJ is written as the variant “㝵” (see Table 3, §18). This form also appears in both Nakakura 23 and the NMC Scroll (see Table 4, § 7). Notably, the Taishō Canon includes a textual note indicating that in the Shōgozō (聖語藏), “礙” is written as “閡” (he), which shares the same meaning of “obstruction”. Moreover, in the term “unobstructed” (wu ai), the second character appears in three different forms in Nanatsu-dera’s RXWJ: “礙” (ai), “㝵” (ai), and “閡” (he).

Similarly, the characters “痰” (tan) and “𤸌” (yin) in the Taishō edition are written as “” and “” in Nanatsu-dera’s RXWJ (Table 3, § 19). The second character “” matches the form found in both Nakakura 23 and the NMC Scroll (see Table 4, § 8). Through comparison, it can be observed that Nanatsu-dera’s RXWJ renders “” as a variant derived from the graph “澹” (dan). This character originally refers to the undulating motion of water and conveys meanings such as “serenity” or “detachment”. It is also interchangeable with “淡” (dan), which carries a similar semantic range. Thus, the difference between “痰” in the Taishō edition and “澹” in the manuscripts is not the result of a scribal error but rather reflects these characters’ phonetic and semantic interchangeability. These variants illustrate the flexibility of character usage in manuscript traditions, where phonetic and semantic similarities often influenced scribal choices.

In the comparison of the three manuscripts, several notable differences emerge, reflecting distinct scribal traditions. Certain graphs in Nanatsu-dera’s RXWJ align with regularized nonstandard characters (kaihua suzi 楷化俗字), clearly diverging from both Nakakura 23 and the NMC Scroll, as demonstrated in

Table 5. Graphic forms in three manuscripts: scribal deviations.

	§ 1	§ 2	§ 3	§ 4	§ 5	§ 6	§ 7
Taishō edition	最	惱	濕	亦	體	熱	觸
Nanatsu-dera’s RXWJ
Nakakura 23
NMC Scroll

, § 1–4. For example, in Nakakura 23 and the NMC Scroll, “惱” (nao) is written as “惚”, whereas in Nanatsu-dera’s RXWJ, it is rendered as “”.14 And “亦” (yi) appears in Nanatsu-dera’s RXWJ in both forms: “” and “”. Conversely, other instances show Nanatsu-dera’s RXWJ closely resembling Nakakura 23 while differing from the NMC Scroll, as illustrated in Table 5, § 5–7. Given scholarly assessments that Nakakura 23 embodies a Southern Dynasty writing style and the NMC Scroll represents a Northern Dynasty style, these observations open up the possibility that the textual basis for Nanatsu-dera’s RXWJ likely derives from a Southern Dynasty manuscript tradition.

A comparative analysis of the scribal conventions found in Nanatsu-dera’s RXWJ and the BRXP fragments reveals that while the manuscript preserves certain character forms indicative of an earlier textual layer, it also exhibits notable modifications. These features imply that Nanatsu-dera’s RXWJ is not a direct copy but rather a product of an evolving textual tradition shaped by successive stages of transmission and adaptation. While some similarities point to a traceable connection with earlier editions, the differences underscore the manuscript’s role within a broader, multi-layered process of preservation, revision, and cultural exchange.

5. Conclusions

The Nanatsu-dera manuscript, transcribed by Japanese monks, serves as an important medium for understanding cultural transmission in medieval East Asia. It exhibits significant similarities in textual style to Dunhuang manuscripts, indicating a shared cultural and textual heritage. Since the Dunhuang version survives only in fragmentary form, the Nanatsu-dera manuscript offers a more comprehensive basis for comparative analysis. Notably, manuscript fragments from this period suggest that the Fifty-Fascicle edition was the predominant format in which the text was circulated.

The presence of variant characters within the manuscript underscores the fluidity of phonetic and semantic interchangeability in the transmission of texts. These variations offer valuable insight into how scribal conventions influenced the evolution of Buddhist manuscripts in medieval East Asia. Some discrepancies arise from the use of nonstandard characters (suzi) while others can be attributed to copying errors caused by the visual similarity between characters and their colloquial counterparts. For instance, Kimura’s annotations note that the Taishō Canon records the characters “光” (guang), “先” (xian), and “充” (chong). In contrast, the manuscript renders them as “無” (wu), likely due to the misinterpretation of the variant form “无” (wu), which closely resembles them. Similarly, the phrase “無礙” (wu ai, unobstructed) appears in the manuscript as “無開” (wu kai), an error that likely resulted from the substitution of “開” (kai) or “閏” (run) for “閡” (he), given the interchangeability between “礙” (ai) and “閡” (he). A particularly revealing example appears in verse “其懷惡心者 不覩如來身” (qi huai e xin zhe, bu du rulai shen, “those who harbor an evil mind will not behold the Tathāgata”; T 278, 618c15). In the manuscript, the character “惡” (e, “evil”) is mistakenly rendered as “慈” (ci, “compassion”), resulting in the phrase “其壞慈心者” (qi huai ci xin zhe, “the ones whose compassionate mind is ruined”; Kimura 1999, p. 596), which dramatically shifts the intended meaning. This alteration likely stems from the visual similarity between the graph for “惡” observed in Nakakura 23 () and the NMC Scroll () and the forms of “慈”, written as “” in Nanatsu-dera’s RXWJ. In other words, although the manuscript contains omissions and copying errors, it also indicates that the scribe required a certain degree of interpretive competence or familiarity with the text in order to accurately understand and reproduce its content.

In conclusion, the manuscript appears to have undergone three distinct stages of transmission. First, it originated from the Fifty-Fascicle edition that was in circulation in China, where a single section was extracted and copied. This manuscript then became a base text and continued to circulate in Japan. Finally, it was transcribed again and underwent preliminary collation, giving rise to a new manuscript. This multi-layered process of textual transmission reflects the complex interplay of preservation, adaptation, and cultural exchange. This case study also suggests an approach that may be applied to other texts in the Nanatsu-dera collection, especially those lacking complete canonical parallels. Future research might extend this method to illuminate broader trends in manuscript transmission and scribal variation across the East Asian Buddhist world.